Podcast Summary: How I AI – "How this PM uses MCPs to automate his meeting prep, CRM updates, and customer feedback synthesis"
Guest: Reid Robinson (Zapier) | Host: Claire Vo
Date: February 2, 2026

Episode Overview

In this episode, Claire Vo interviews Reid Robinson, a Product Manager on AI at Zapier, about practical, tactical ways to apply AI—specifically with Multi-Channel Processors (MCPs)—to automate and elevate tasks like meeting preparation, CRM updates, and synthesizing customer feedback. Reid shares screens and discusses how he configures his workflow, offering actionable advice for listeners wanting to supercharge their productivity and move beyond tedious manual work. The episode is full of detailed examples, real-life tips, and creative applications of AI for both work and personal life.

Main Topics and Key Insights

Clarifying MCPs and Why They Matter

[00:00–05:00]

• MCPs act like "app integrations for your AI tools." They allow you to connect the AI tools you love (Claude, ChatGPT, Cursor, etc.) to the business apps you rely on (Google Calendar, Slack, CRM, etc.).
• Reid: “Don’t think about the word. It really just is like app integrations for your AI tools. You can create these collections of tools from all the apps you use and give them access to Claude, to ChatGPT, to Cursor…” [00:05]
• Main use cases:
  1. Give AI access to internal/external app knowledge.
  2. Let AI actually perform real-world actions in those apps.
• Claire highlights the challenge of MCP branding—users need approachable language to understand these concepts.

Inside Zapier’s Approach to MCPs

[05:28–09:08]

• Zapier acts as a “platform for creating MCP servers”—not just connectors, but customizable bundles for specific agents or workflows.
• You can configure granular permissions, e.g. allow Claude access to just certain Coda docs or specific Slack channels.
• The flexibility allows users to tailor experiences for multiple agents (Claude, ChatGPT, Cursor, etc.) with a single URL.

Quote:
“For my use within Claude, I really am using it for particular documents and particular sheets…with Evernote, I want to restrict it to writing to certain notebooks.” —Reid [06:13]

Use Case 1: Meeting Prep & CRM Automation

[09:08–17:02]

• Reid describes setting up Claude Projects for customer research and CRM work.
  • Uses MCPs to sequence tool access, ensuring the right database is queried, notes are logged correctly, and lookups are comprehensive.
  • Advanced: Embeds detailed, workflow-specific instructions into Claude’s projects to guide tool usage.

Quote:
“In Claude projects you can provide very, like, detailed instructions for use cases … I have one that's all about the way I like logging and looking up data from CRM, from our CRM for things.” —Reid [09:08]

• Massive time-saver for customer-facing teams who “hate” updating CRMs.
  • Claude, with the right MCP configuration, can look up meeting participants, grab details from company usage and sales records, and summarize it all for daily planning.

Quote:
“...keeping good customer records, whether it’s for a sales use case, a research use case, whatever is like really tedious. And there are actually amazing MCPs out there to do this.” —Claire [10:20]

Deterministic vs. Agentic Workflows

[15:26–19:12]

• Claire contrasts “deterministic” workflow builders (step-by-step) with agentic, instruction-driven processes (via AI agents like Claude).
  • Benefit of agents: “In natural language, describe that flow.”
• Discussion of reliability—agentic is better for flexibility and context; deterministic is better for long, complex, or timing-sensitive processes.

Quote:
“It is hard to break that muscle memory of like, this is a, you know, a deterministic workflow versus an instructive agent, even if it has access to the same tools and can do the same things.” —Claire [15:26]

• Enterprise adoption: Automating tool access and workflows so new employees get the right tools for their roles by default.

Use Case 2: Automated Customer Feedback Synthesis

[25:38–29:16]

• AI-driven bots analyze support tickets and chat logs, propose FAQ entries, and (after human review) update internal knowledge bases.
  • Systematic, high-quality improvement of help docs, keeping internal/external knowledge current without manual effort.

Quote:
“Every time there is a closed support ticket or if there’s a finished chatbot transcript, it analyzes the conversation…If not, please propose an entry…if I approve it, it goes over to a different database which is the one that the bot is actually using. So a really nice way…to like rapidly iterate and keep those things up to date.” —Reid [28:00]

Use Case 3: Meeting Prep with Data Integration

[20:33–24:09]

• Workflow involves integrating Databricks lookups into prep flows—automatically fetching research on meeting participants and appending it to the relevant Coda page.
  • For large companies, getting context into meetings is crucial, especially if bookings come from many channels.

Quote:
“It goes out, fetches that…research lookup which takes time and then…creates the, or appends it technically to the coda page for that customer interview. And so this is really helpful for me…when I’m going into my meetings, I get things like this…so I can actually see like, oh, they did use it and get some really crisp things to walk into the meeting knowing.” —Reid [21:45]

• On choosing the right model for the job (Gemini’s file handling):
  “The output from our data team is actually to date a PDF…so it works very well.” —Reid [24:09]

The Power of AI as an “Always-On Team”

[29:16–30:55]

• Claire encourages listeners to imagine what work could look like with “the perfect team with infinite time”—what would get done? AI makes that level of thorough, proactive effort possible, especially for support and customer experience.

Personal & Unexpected MCP Use Cases

Lightning round [33:27–36:53]

• Family Calendar Automation: Taking a photo of a physical family calendar → interpreted and uploaded to Google Calendar with drive time considered.
  • “Now what I do is like, occasionally, I just take a picture of the physical calendar and then it…does all of it. And I love that.” —Reid [33:44]
• AI-Generated Kids Songs with Suno & Claude: Turning daily experiences (with requisite “poop and fart jokes”) into songs via AI, not just for fun, but as prompting education for kids.
  • “My son has listened to this at least on Suno alone 14 times...and they've made songs for each other and it's really fun...one of the girls is like 10...she's been learning about prompting through this…” [34:50]
• AI for Partner Support (Job Prep): Used Notebook LM to prep his wife for job interviews, generating personalized briefings and audio overviews so she was always the “most informed applicant.”
  • “She listened to all of these before and she, like, constantly got feedback throughout the process that she was like the most informed applicant. She clearly understood the space…” [37:42]

Notable Quotes & Memorable Moments

• On abstracting MCP complexity:
  “Think about if I could give my favorite AI chat client or IDE or whatever access to a bunch of tools to do things for me, what would I want them to do? And then go hunt for an MCP that does that thing.”
  —Claire [03:32]

• On the shift from workflows to agents:
  “It is hard to break that muscle memory of like, this is a...deterministic workflow versus an instructive agent, even if it has the same access to the same tools...”
  —Claire [15:26]

• On AI-driven knowledge base upkeep:
  “A really nice way that I have found...to like rapidly iterate and keep those things up to date so that users are just getting their answers faster.”
  —Reid [29:00]

• Childlike joy with AI songwriting:
  “My son insisted that it have poop and fart jokes in it as well...my son has listened to this at least on Suno alone 14 times...”
  —Reid [34:37]

Timestamps for Important Segments

• 00:00–05:00 – What are MCPs and why are they so misunderstood?
• 05:28–09:08 – How Zapier lets users build custom MCP bundles and streamline tool access.
• 09:08–17:02 – Real-world flows: Automating CRM updates and meeting notes with Claude + detailed project instructions.
• 15:26–19:12 – Deterministic vs. agentic workflows and the reliability debate.
• 20:33–24:09 – Integrating databases and picking the best AI model for the task (Gemini/PDFs).
• 25:38–29:16 – Using AI for feedback synthesis and FAQ upkeep.
• 33:27–36:53 – Lightning round: Family calendar automation, songs for kids, and AI-aided job searches.
• 37:22–38:35 – Using Notebook LM for personalized partner support.

Tone & Final Thoughts

Both Claire and Reid keep the discussion pragmatic yet playful, offering up concrete examples (“If you could run ChatGPT in your sleep, what would you do?”) and lively, relatable anecdotes. The through-line: AI is as much about making work better as it is about enabling richer, more joyful lives—at work, at home, and everywhere in between.

Where to Find More

• Connect with Reid: LinkedIn / Reid Robinson
• Try Zapier MCP: zapier.com/mcp
• More episodes, resources, and show notes: howiaipod.com

End of Summary